"In the R&D and application of AI technology, we need to pay attention to both training and inference. Training is the foundation of AI technology creation capabilities, and inference is the key to monetization. ”
Text: Technode
作者｜黄尘排版｜Miziko
This article is expected to take 8 minutes to read
The growing adoption of AI applications and the rapid development of generative AI have deepened the connection between the mobile industry and other sectors, and building a robust infrastructure to support AI development is a critical task today. PricewaterhouseCoopers previously released a report that artificial intelligence will contribute $15.7 trillion to the global economy by 2030. In this dividend cake, the technology will drive China's GDP growth by 26%.
As the fascination with generative AI has passed, the industry is clamoring for AI applications and services that can be implemented in real terms. Businesses and consumers are looking for AI solutions that directly solve real-world problems, increase efficiency, and enhance experiences. This shift marks an important turning point in the development of AI – from proof-of-concept and experimental exploration, to real-world integration and widespread application.
AI is leading the transformation of content interaction and the development of the Internet of Everything
In a rapidly evolving modern society, AI is increasingly becoming a key force driving social progress and innovation. Speaking at MWC 2024, Wang Tao, Executive Director of the Board of Directors and Chairman of the ICT Infrastructure Business Management Board, said, "AI technology is leading another revolution in content interaction through its powerful data processing and learning capabilities. From improving the efficiency of interactions, to improving the quality of interactions, to expanding the number of objects to interact with, AI is driving the explosive growth of mobile communication traffic. ”
For example, generative content will replace traditional retrieval content, and AI assistants will be able to automatically check massive documents and generate images in seconds, greatly improving the efficiency of information acquisition. At the same time, with the development of IPC technology, the production mode of full-modal content will increase the amount of interactive information by more than 100 times. In addition, with more than 100 or even more local small models built into each terminal, the terminal interaction object will expand from a single human-computer interaction to more machine interactions, further promoting the growth of traffic.
The application of AI technology in the automotive field will completely change the way people travel. Connected cars will not only provide more enjoyable and reassuring travel services, but will also disrupt traditional driving modes through intelligent cockpits and autonomous driving technology. The 100-fold increase in on-board traffic and the monthly upload of 100 GB of data in autonomous driving mode will greatly promote the development of the mobile communication industry.
AI technology will greatly expand the boundaries of the Internet of Everything. As machines gain the mind and mobility of content, network connectivity will expand from carbon-based people to silicon-based people, creating tens of billions of new people. At the individual level, everyone will have an AI assistant to respond to various needs in real time, enabling the evolution from tool to partner. At the industry level, smart factories will be equipped with multiple AI brains to drive AI machine employees to go deep into each production link to improve quality and efficiency. It is estimated that by 2030, AI robots will enter more than 80% of factories in developed countries, penetrate more than one-third of the world's production jobs, and co-produce and work with humans.
The adoption of on-device AI is accelerating
The development of device-side AI has been a hot topic in the field of science and technology in recent years, which represents the deep integration and application of artificial intelligence technology on terminal devices.
Specifically, device-side AI is the operation and processing of AI technology on terminal devices. Compared with traditional cloud-side AI, device-side AI delegates data processing and analysis capabilities to terminal devices, such as smartphones and smart home devices. This shift has multiple implications: first, it reduces the dependence on cloud servers and increases the speed and efficiency of data processing; Second, it enhances the security of the data because user data is processed locally, reducing the risk of data breaches; Finally, it promotes the popularization and application of smart devices, enabling more devices to have intelligent functions.
The development of on-device AI is inseparable from the support of a series of core technologies. Among them, machine learning algorithms are the foundation of device-side AI, which enables terminal devices to learn and improve autonomously. The deep learning framework provides efficient model training and inference capabilities; The neural network optimization technology optimizes the computing power of the device, which improves the operation efficiency of the model. Hardware acceleration technology uses hardware resources to accelerate the computing process, further improving the performance of AI on the device.
The application scenarios of on-device AI are wide and diverse. In the field of smartphones, device-side AI can implement functions such as face recognition, voice recognition, and natural language processing, providing users with a more convenient and intelligent interactive experience. In the field of smart home, device-side AI can realize the intelligent control and management of home devices, improving the comfort and convenience of family life. In the field of autonomous driving, end-to-end AI can realize autonomous navigation and decision-making functions of vehicles and improve traffic safety. In the medical field, device-side AI can assist doctors in disease diagnosis and treatment plan formulation, improving the quality and efficiency of medical care.
Apple Intelligence, which Apple recently disclosed at WWDC 2024, reconstructs the underlying operating system of iOS 18 and deeply integrates AI into Apple's entire product line of smart devices. Industry insiders believe that Apple's strategy is similar to the four-layer AI architecture proposed by Honor, in this regard, Honor CEO Zhao Ming said: "All manufacturers will definitely follow our path, because this is the correct development direction and strategy of mobile phone manufacturers in AI...... In the future, the road of device-side AI will definitely become wider and wider, not only to put the capabilities of network-side AI on hardware, but also to really use AI to reconstruct the operating system and combine it with various underlying hardware. ”
Although end-to-device AI has broad development prospects and huge application potential, it also faces some challenges and opportunities. Challenges include hardware performance limitations, data privacy and security protection, model complexity, and the balance of computing resources. To solve these problems, we need to continuously explore new technologies and methods to improve the performance and security of on-device AI. At the same time, with the continuous development of technologies such as 5G and the Internet of Things, on-device AI has also ushered in new development opportunities. These technologies provide more abundant data sources and more efficient communication methods for device-side AI, enabling device-side AI to better play its role in various fields and bring more convenient and intelligent experiences to people.
AI challenges and solutions
With the rapid development of artificial intelligence technology, we are ushering in an era of unprecedented change. However, the development of AI technology has not been without its challenges. In the process of AI technology development, there are currently three major challenges: one is the supply bottleneck of computing power, the second is the accuracy of AI in key scenarios, and the third is the challenge of building a commercial closed loop of AI technology. These challenges need to be faced together and solutions found.
In order to overcome these challenges, Xu Ziyang, president and executive director of ZTE, put forward a number of ideas. Specifically, by building a larger-scale computing power cluster, the effective supply of computing power is realized. This requires strengthening the connection and collaboration at all levels such as chips, servers, and data centers to jointly build an efficient and reliable computing network. In the R&D and application of AI technology, it is necessary to pay attention to both training and inference. Training is the foundation of AI technology creation capabilities, and inference is the key to monetization. Only by paying equal attention to training and reasoning can we ensure that AI technology can play a greater role in the real economy.
Of course, the development of AI technology is also inseparable from the joint efforts of the entire industry chain. Therefore, it is also necessary to strengthen cooperation between the upstream and downstream of the industrial chain to jointly promote the research and development, application and promotion of AI technology. Through cooperation, upstream and downstream enterprises can share resources, reduce costs, improve efficiency, and jointly promote the development of AI technology.
Zhang Yu, CTO and Chief AI Engineer of Intel China's Network and Edge Business Unit, said: "AI requires the full cooperation of the entire industry chain, it requires companies like Zhipu AI and Baidu to build some services such as large models, companies like Lenovo to enable these services to land in specific hardware, and chip companies like Intel to provide underlying computing network support." ”
epilogue
Looking ahead, AI technology will continue to unleash its potential in various fields, not only contributing to the huge growth momentum of China and the global economy, but also bringing far-reaching changes to people's work and life. With the continuous maturity of technology and the deepening of applications, a new era of AI-driven intelligence will come to us.
This article is an original article by TechNode reporters, and may not be reprinted without authorization.
- - - - - - - - END - - - - - - - -
Interactive topics
What are your expectations for AI technology?
Come and leave your opinion in the comment section!
*If you want to get industry information and share your experience with like-minded technology enthusiasts, then scan the QR code to add "Movedian" to join the group chat! There are more irregular benefits in the group!
Wonderful article is worth recommending
One-click forwarding, poke and watch